|
--- |
|
language: |
|
- it |
|
--- |
|
## GPT-ita-fdi_lega🇮🇹 |
|
|
|
|
|
Finetune of an Italian version of gpt-2 ([GePpeTto](https://huggingface.co/LorenzoDeMattei/GePpeTto)) trained on tweets of politicians from the far right Italian parties FDI and Lega. |
|
|
|
|
|
## Finetuning corpus |
|
|
|
The model was finetuned over a private dataset of tweets from italian politicians. The tweets were collected between 2021 and 2022 from the Twitter accounts of all the "FDI" and "Lega" members of the Italian Parliament. |
|
In the end, the finetuning was conducted over a corpus of ~40K tweets |
|
|
|
## Uses |
|
|
|
By giving the model a few Italian words to start from, the model can generate a tweet in the style of far right Italian politicians. Try it out [here](https://huggingface.co/spaces/ruggsea/demo_gpt-ita-fdi_lega) |
|
|
|
|
|
## Bias, Risks, and Limitations |
|
|
|
Compared to the base italian gpt-2 model, this model could generate more hateful or toxic content and exhibit bias, in line with the training corpus. |
|
|
|
### Recommendations |
|
|
|
|
|
Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. |
|
|
|
## How to Get Started with the Model |
|
|
|
Use the code below to get started with the model. |
|
|
|
|
|
``` |
|
from transformers import GPT2Tokenizer, GPT2Model |
|
|
|
model = GPT2Model.from_pretrained('ruggsea/gpt-ita-fdi_lega') |
|
tokenizer = GPT2Tokenizer.from_pretrained( |
|
'ruggsea/gpt-ita-fdi_lega', |
|
) |
|
``` |
|
|